Homebrewing an XYZ CNC Router

  Quoted Text Here  

The logical conclsion of which would be a CNC machine with no steppers or servos as we know them now, but withmotion determined only be the detection of the current cutting point, together with a machine vision system to assess the geometry and positioning of whatever has been cut so far (To take into account warping and toughness of the article being machined)

On Wed, 26 Aug 2009 10:53:06 +0100, "Phil O. Sopher"
  Quoted Text Here  
When you're done writing the software & developing the control system to do all this, let me know...but I won't hold my breath in the meantime ;-)
Regards, Tony

On Wed, 26 Aug 2009 09:28:18 +0100, "Phil O. Sopher"
I believe Mach 3 has the ability to correct for non-linearity in the leadscrews, but that's about it.
Regards, Tony

  Quoted Text Here  
Right, I'm not pro engineer or anything close, just an interested lurker, but I don't understand something with this discussion that seems like an obvious gotcha.
Your thinking homebrew router, right? Yes bits and bobs will flex about in all sorts of various ways. With a printer or scanner, there are no forces opposing the heads really. But a router has a rotating tool getting messed about in all sorts of dynamic ways. Its rotating, it gets pushed up down left right, the tool gets hot, different forces apply as it changes direction while rotating, and probably many other things to worry about. I imagine that all gets very hard to model. So in general , I would have thought that when flexing happens in machines, the solution is a mechanical one and not a software one. Build the bugger stronger!!!
With all that in mind, I don't quite see who would write a program doing what you suggest , when the simple solution is to build the thing properly and address flexing with a mechanical solution. I don't know why, but that seems intuitively the obviously correct thing to do: Build it more ridged. It has to be easier to do that than model out all the dynamics involved.
So, I'm thinking that since such software does not exist now, then there must be a good reason why not. If it were possible to build a weak machine and simply compensate for it in software, every one would do it. I'd have my home brew MDF CNC thing milling out baby engine blocks!!
Makes me think its either not possible since the maths are on a metrological level, or not economic in some way. Perhaps it can be done, but we'd need met office computers to do the calculations!!!
Anyway, see what I mean? Have I missed something painfully obvious?
AC

  Quoted Text Here  
Correcting for static misalignment seems far more feasible than trying to compensate for cutting strain. However, there's still a problem.
If you distort the cutting path to allow for a twisted base, the cut surface would then, presumably, have the intended shape. But the object overall won't : the base will follow the twisted bed, or approximate it. In the extreme case, you can imagine a perfect cut surface but a bulge underneath. Depending on the material, that might bulge the top surface if bolted down to a truly flat base.
You may be able to remount the piece and flatten the bottom using the same compensation. That's probably significantly harder to compensate as it will inevitably be less well supported.
-adrian

Assuming we have a clever group that can provide the software and enough computing power etc.etc. How the hell do we calibrate this homebuilt, less than ridgid device with nonlinear axes?
Henry

  Quoted Text Here  
The formula will require a constant on which to base the nonlinearlessthanrigid factor on, Nonlinearlessthanrigid is actually a word according to Ower Gert who always beats me at scrabble.
After countless sleepless nights and talking to all forms of drunken unemployed missfits at the local pub, [ bastard f a job but someone has to do it ] I have come to the conclusion that this rare figure is a culmination of of a time / space equation so as a starting point we should use the time taken to strain 3 litres of Asda fine piccalilli though a pair of 30 denier tights [ not whilst being worn though ]
John S.

John S wrote:
  Quoted Text Here  
:)
Actually, it's mostly quite doable, in theory at least.
Take three points (not in line) somewhere above the cutting point, and fix scales between them and the cutting point. These points form a triangle, which is of mathematical necessity exactly in a plane, indeed they define the plane.
From the outputs of the scales you can work out a position for the cutting point, then it's fairly simple to do a feedback loop so that the cutter is in the correct position.
With good enough scales and very fast and powerful feedback you might be able to do accuracy to a few microns, or even better.
Actual design is left as an exercise for the reader.
-- Peter Fairbrother

On Tue, 01 Sep 2009 12:05:22 +0100, Peter Fairbrother
  Quoted Text Here  
This only works if the three points that anchor the ends of the scales are fixed relative to the workpiece. If flexing of the machine results in the workpiece moving relative to the 3 points, or if the workpiece itself flexes (which it does, of course) then all bets are off.
Regards, Tony

>>Actually, it's mostly quite doable, in theory at least.
  Quoted Text Here  
Plus, if I've understood the suggestion correctly, while you may be able to establish some reference points- between them you will need to make some assumption about how to interpolate the correction(s) and that is before further flexing occurs.
I've done some modelling of errors in 3D due to flexing in the past (for attitude control purposes) but the model was only as good as the "fit" you applied for interpolation. In the application, the flexing was very limited (in comparison to the overall dimensions of the ship) and was (surprisingly) repeatable. As a result, a model was possible, with a "look up table" to apply corrections based on ship dynamics.
If the flexing in the work piece and machine bed was reasonably repeatable (which seems unlikely), I suspect it could be modelled. Not a trivial task but an interesting one.
Brian

wrote:
  Quoted Text Here  
If you are taking chunks out of the workpiece (which you are, otherwise there's no machining occurring), then any flexing of the workpiece is subject to continuous change as the machining progresses. Very definitely not a trivial task to figure out *how* it is going to change, depending on position of cut, depth of cut, shape of cutter,...
Regards, Tony

  Quoted Text Here  
Exactly. May be in a production environment you could model a particular job but, for one offs, it wouldn't seem practical. Having said that, if someone can do it, I'd like to see the sums ;-)
Brian

Tony Jeffree wrote:
  Quoted Text Here  
Y'rright, of course - the points should be fixed to the workpiece, not somewhere else. But even that doesn't compensate for workpiece flexing.
No coffee :(
However if we can ignore workpiece flexing, then fixing the points to the table is pretty good
-- Peter Fairbrother

  Quoted Text Here  
To be honest, it's not something I'd yet given a lot of thought to. (But I didn't start this off by proposing a less-than-rigid device; I was contemplating dealing with errors that might arise despite one's best intentions as to solid and sound construction)
However, this is a discussion group, so I suppose that it's a point worth discussing.
I'd try to conceive of something akin to the parallel bar used to assess alignment of lathes. Perhaps a hardened dead square block against which one could traverse the chuck, with just a scriber in it, and then measure the growing gap as the scribe moved along the block?
This is the sort of thing where soultions would present themselves as the result of hands-on experience rather than as an up-front thought experiment.

  Quoted Text Here  
You always manage to hook 'em don't you?
:)
Steve